Artificial intelligence has become one of the most talked-about technologies in the JD Edwards community, but moving beyond experimentation remains a challenge for many organizations. During his BLUEPRINT 4D 2026 session, “Take your JDE AI Solution from POC to Production,” Vevin Kumar shared practical lessons from Martin Marietta’s journey implementing AI-driven solutions in a production JD Edwards environment. Rather than focusing solely on exciting demonstrations, Kumar explored what it actually takes to deliver secure, scalable, and measurable AI solutions that provide business value.
With operations spanning more than 500 locations across North America and thousands of JD Edwards users, Martin Marietta has a significant opportunity to leverage AI. The company’s experience demonstrates that while building a proof of concept (POC) is often straightforward, transforming that prototype into a production-ready solution requires a completely different level of planning and execution.
The Difference Between a POC and Production
One of Kumar’s central messages was that successful AI initiatives must be designed with production requirements in mind from the beginning. A proof of concept only needs to demonstrate value and generate excitement. Production systems must withstand security reviews, scale to thousands of users, and deliver consistent results.
Martin Marietta’s first major AI initiative was a conversational assistant that allowed users to interact with JD Edwards using natural language. Employees could ask questions such as whether they had orders awaiting approval, identify top customers, or review business performance metrics. Behind the scenes, the solution used JD Edwards Orchestrations to retrieve information and present it through a user-friendly chat interface.
The original POC succeeded in demonstrating the concept, but Kumar quickly discovered that the architecture needed significant refinement before deployment. Direct connections between users and AI services were not suitable for enterprise use. The team introduced API gateways, scalable cloud-hosted services, authentication layers, and role-based access controls to create a secure and manageable architecture.
The lesson was clear: production AI requires enterprise-grade architecture.
Building Secure and Scalable AI for JD Edwards
To support thousands of potential users, Martin Marietta implemented a layered architecture hosted in Microsoft Azure. The solution incorporated load balancing, API management, encrypted communication, authentication tokens, and security controls at every level. AI requests were routed through approved orchestrations rather than allowing direct database access, reducing the risk of inaccurate or unauthorized responses.
A particularly interesting component was the use of a vector store containing customer, product, and plant master data. This allowed users to ask questions using familiar business terminology instead of JD Edwards identifiers. For example, a salesperson could ask about bicycle sales rather than entering a specific item number. The AI could translate the business language into the identifiers required by JD Edwards and retrieve the correct information.
The approach also helped reduce AI hallucinations by ensuring that all business data ultimately came from JD Edwards through controlled orchestration services. While not completely eliminating errors, the architecture provided a high degree of reliability and control.
Managing Cost Without Breaking the Budget
Another common concern surrounding AI initiatives is cost. Kumar emphasized that organizations do not need massive budgets to begin experimenting with AI.
Martin Marietta’s chatbot architecture incurred relatively modest expenses. Cloud hosting costs were low, vector storage costs were negligible, and model expenses were optimized by moving from larger AI models to smaller, more efficient versions that delivered comparable results for their use case. By carefully monitoring usage and selecting the appropriate model, organizations can control spending while still delivering value.
The key takeaway was that successful AI projects are often less about technology costs and more about disciplined architecture, governance, and usage management.
Using AI to Accelerate JD Edwards Upgrades
Beyond chatbots, Kumar shared a practical example of AI accelerating JD Edwards upgrade activities.
During a Tools Release upgrade, Martin Marietta discovered that several custom-built JET pages became incompatible due to underlying library changes. Manually fixing approximately 20 customized pages would have required significant developer effort and specialized expertise. Instead, the team leveraged AI coding tools to compare Oracle’s older and newer JET page implementations, identify patterns, and automatically update custom code.
The result was impressive. AI modified roughly 1,000 lines of code across dozens of files, reducing what could have been weeks of manual effort to a single afternoon. Rather than replacing developer expertise, AI amplified it by handling repetitive pattern recognition and code transformation tasks.
Faster Troubleshooting Through AI Analysis
AI also proved valuable in production support scenarios.
Martin Marietta encountered recurring zombie processes and kernel crashes within its JD Edwards environment. Traditionally, analyzing memory dumps and identifying root causes would require extensive investigation by technical teams. Instead, the organization provided multiple dump files to an AI model and asked it to identify common patterns.
Within minutes, the AI identified a specific database trigger and related business function that appeared consistently across all crash logs. The analysis accurately traced the call stack, identified user actions triggering the issue, and highlighted the recursive logic responsible for database locking. After implementing a temporary fix, the zombie processes immediately stopped occurring.
For JD Edwards technical teams, this represents a compelling use case: AI can dramatically reduce the time required to interpret complex logs and identify root causes.
Reimagining the JD Edwards User Experience
Perhaps Kumar’s most forward-looking example involved modernizing JD Edwards user interfaces.
Using AI-assisted development and modern frameworks such as React, Martin Marietta redesigned highly customized JD Edwards applications while still leveraging existing JD Edwards infrastructure and AIS services. Instead of presenting users with multiple grids, scrolling, and complex navigation paths, the new interfaces consolidated information into intuitive dashboards, cards, filters, and drill-down experiences.
The redesigned applications were responsive, mobile-friendly, and easier to navigate than traditional forms. Users could access information more efficiently while maintaining seamless integration with standard JD Edwards applications and security controls.
Most notably, AI significantly accelerated development. Once the foundational framework was established, new applications could be generated in hours rather than weeks, dramatically improving development velocity.
Focus on ROI, Not Just Innovation
Kumar concluded with a reminder that executive leadership ultimately cares about measurable outcomes. Successful AI projects require clear metrics demonstrating value, whether through time savings, productivity improvements, cost reductions, or enhanced user experiences. His team quantified improvements by measuring application usage, estimating time savings per interaction, and translating those gains into business impact.
The broader message for JD Edwards organizations is that AI should not be viewed as a standalone technology initiative. It should be treated as a business improvement strategy supported by strong architecture, measurable outcomes, and a clear path from experimentation to production.
As AI capabilities continue to evolve, organizations that combine solid JD Edwards expertise with practical AI adoption strategies will be best positioned to unlock meaningful value from both technologies.
Want more?
Explore more content and resources to help you get the most from your Oracle investment:
- Visit the Quest Learn Library for blogs, how-to demos, and on-demand sessions.
- Connect with peers in one of our Quest JD Edwards Community User Groups to swap stories, ask questions, and share tips with other users facing the same challenges.
- Bookmark Learn JDE for updates from JD Edwards development.
Not a Quest member yet? Join today and tap into the ultimate Oracle customer network.